International Journal of Computer Applications (0975 – 8887) Volume 104 – No.1, October 2014 14 Domain Specific Knowledge based Machine Learning for Flower Classification using Soft Computing Rajesh S. Sarkate Dept. of CS & IT MGM’s College of CS & IT, Nanded Prakash B. Khanale, PhD Dept. of Computer Science DSM College Parbhani ABSTRACT Machine learning is ability of the machine to capture the data, analyze it and make decision as like human being perform in everyday life. Sometimes it also referred as pattern recognition or classification. With use of latest computing technology and soft computing, it is now possible for machines to act with intelligence as human. Various industries like automobile, medical diagnosis are using machines for fast and accurate data capturing and analyzing. Our study is conducted to justify these machine abilities in the field of floriculture. A knowledge base in flower domain is used for the Intra-class sorting purpose. The flower sorter is designed to capture the flower images and with artificial neural network classifier, the images are classified in four flower color classes. In the study, supervised learning algorithm is used for machine learning. The classification accuracy of flower sorter is found to be 98%. General Terms Pattern Recognition, Horticulture, Neural Network. Keywords Machine Learning; Classification; Flower; Sorting; ANN; Knowledge base. 1. INTRODUCTION There was a time when experts had doubts in human being replaced by machines as the capabilities and performance especially in the sagacity of machines were debatable. Nevertheless, studies conducted in this domain have proved not only the capabilities but also the perfection of machines over human limitations. In the current study has investigated an approach of machine learning, specifically, with the superior computational ability of segmentation and for the judgment of the color wise discrimination in flowers. This is possible with superior computer storage and computation for extracting statistical features and for locating objects in a high-dimensional feature space. In past, flowers from different varieties were classified by researchers based on color, texture and shape as features. The present study concentrates of Intra-class separation of single flower specie i.e. Gerbera jamesonii with limited feature i.e. color as shape and texture remain same. It is possible that in the domain of flower recognition, current study could help in commercial development of human unaided machine automated flower sorter system. The knowledge base that uses priory information of specific domain guides to interpret and learn the machine about the objects of interest i.e. flowers. In addition, the proposed method can be used in applications that require human free interface to other agriculture products. Computer vision technology is widely used in Agriculture[1] for pest or disease detection[2][3][4], lesion estimation[5], grading[6][7]. In this flower recognition system, a domain- specific knowledge is base for machine learning. The key to efficient classification is the flower colors that allow the, like human, machine to use the features throughout the classification process. [8] Gerbera (Gerbera jamesonii) is top in the list of most popular commercial cut flowers in the world and according to the global trends in Floriculture; it comes in top ten among cut flowers[9]. Throughout the year, Gerbera flower are in demand in both domestic and export markets due to attractive blooms that are suitable for any type of floral arrangements and are available in different shades and hues. The cut flowers have long vase-life reasoning premium market prices. Due to changes in social and cultural life style of people, cut flowers have found an important place in various social functions and daily activities. In many plants, flowering is conspicuous in the field. Flowers are displayed prominently at the top of the canopy as in crops such as Gerbera and they appear perpendicular to the stalk. While flowering is an important developmental stage in most crops, enumerating flowers is labor intensive, especially when flowers need to be counted on a daily basis. Human intervention in counting process may lead to the physical contact with the plants resulting in damage to plants. The development of low-cost color digital cameras that use charge coupled device (CCD) arrays to capture images offer a potential method for measuring flowering in plants with appropriate architecture and flower color.[10] 2. MATERIALS AND TOOLS The different flowers samples used in the present work are collected from main flower market in Nanded district of Maharashtra state in India for the production year 2013. The images are acquired with a color Digital Camera, and then transferred to computer system with core2deo @2.4 GHz. The camera is mounted on a stand with a facility for vertical movement to fine tune the orthogonal distance of the camera from the flower samples in a properly illuminated chamber. The images are illuminated with light source of 40W, 230 V fit to the test table at an angle of 45 0 from the camera. Camera is calibrated to get petal part that decides the flower color. A database of 50 flowers from 4 classes with the help of digital camera is developed for the study. The pictures are deliberately taken with different resolutions i.e. 320 x 240 pixels, 1600 x 1200 and 2592 x 1944, to check resolution